iMac 2020 fluctuating Geekbench scores?

[mlykke]

I'm deep in the decision process, trying to decide between the i7 and the i9, and also trying to decide upon nano-texture or not.
So I've gone through just about every single review video and post on this forum with benchmarks etc. of the new iMac. But the Geekbench scores seems to vary quite a lot - More than I recall seeing with previous computers.
One person with the i7 gets a score just below 9000 while another person with the same config doesn't get much more than 7500 - Thats a huge difference in actual performance.
Browsing the Geekbench results also shows anything from 6486 to just above 9184 for the i7 3.8 Ghz. Since the i7 version can only be bought in a few different configs, none of which should have much influence on the single-core/multi-core score, I would expect the score to be within 2-300 points in most tests. Can anybody share some insights on this?

Getting an i7 with gives a score of close to 9000 would be nice, but I would be deeply disappointed of getting one that only gives 7-7500.

 

[wendylou]

Excellent question & observations. I keep canceling and reordering multiple times between the 8-core i7 and 10-core i9 based on Geekbench. The scores are all over the place! What to do?? My latest order is a 10-core i9. I hope it’s a good call. I worry that 8-core i7 could have higher performance... why are the scores dancing around so much between these two processors?

 

[jobinhosyntax]

My guess is people incorrectly installing the RAM... but could be wrong.

This French review shows the i7 outperforming the i9 on the Logic benchmarks. Stating that each core on the i9 clocks around 1.6 while the i7 clocks around 4.5. They also show the i9 scoring only 8000 on Geekbench 5 while most i9s score 10’000. Either their unit is faulty or they installed the RAM incorrectly. What do you think?

Test de l'iMac 27" 2020 : le dernier des Mohicans ! (+ vidéo)

Pressentis pour la WWDC, Apple aura finalement attendu le milieu de l'été pour renouveler ses iMac !

www.mac4ever.com

 

[mlykke]

In theory the i7 should be awesome for even a quite heavy workload and a lot of the results support that. But there is just too many results way below 8500-9000, that worries me a bit.

 

[Whiteman007]

My guess is the lower scores are form reviews that added ram with the Apple ram and are either losing duel channel or getting die clocked to 2667 MHz

 

[mlykke]

I just saw Barefeats have posted benchmarks for the i7 and only getting 7321 in multicore. That is VERY low.

2020 iMac 5K 8-core versus 2017 iMac Pro 8-core

real world speed test results for performance minded Macintosh users

barefeats.com

 

[mlykke]

My guess is the lower scores are form reviews that added ram with the Apple ram and are either losing duel channel or getting die clocked to 2667 MHz

Yeah that was my first thought, but then I started comparing only results with 16, 32 or 64 gb ram to avoid the ones with 40 gb and similar. But still, I'm seeing big differences in the numbers. I would also expect that the person behind Barefeats have done enough benchmarks to know about this, and the results he posted are very low.

 

[KrazyKanuck]

Yeah that was my first thought, but then I started comparing only results with 16, 32 or 64 gb ram to avoid the ones with 40 gb and similar. But still, I'm seeing big differences in the numbers. I would also expect that the person behind Barefeats have done enough benchmarks to know about this, and the results he posted are very low.

I suspect it's a combination of RAM configuration and some benchmarks being performed while the system (Spotlight) is still indexing files. My i7 reliably scores above 1380 SC and 8900 MC with 32GB of aftermarket RAM (four sticks of Crucial 8GB). That is what you should expect.

EDIT: I should add that the scores I cited were achieved without any kind of additional fan control. I'm sure also that one or two users are running Geekbench with their iMacs sitting in refrigerators (or fans blasting, or in front of an AC, or with a vacuum cleaner nozzle stuck to the exhaust 🤪) to get top scores, so you should also be skeptical of the highest scores.

 

[mlykke]

My i7 reliably scores above 1380 SC and 8900 MC with 32GB of aftermarket RAM (four sticks of Crucial 8GB). That is what you should expect.

Thanks for the info - Thats what I'm hoping for. I just don't remember seeing so varying results from previous iMac launches.

 

[KrazyKanuck]

Thanks for the info - Thats what I'm hoping for. I just don't remember seeing so varying results from previous iMac launches.

Agreed. For some reason this iteration seems to be quite particular about mixing RAM. I'm sure that this is contributing to the wildly varying scores.

 

[wendylou]

So does this mean the 10-core i9 should have higher scores than the 8-core i7? I would hate to commit to ordering the i9 if the results are lower than the i7, not to mention it costs $400 more. I’m looking for max power so is it a safe bet that the i9 is the better choice?

 

[KrazyKanuck]

So does this mean the 10-core i9 should have higher scores than the 8-core i7? I would hate to commit to ordering the i9 if the results are lower than the i7, not to mention it costs $400 more. I’m looking for max power so is it a safe bet that the i9 is the better choice?

The multi-core score for the i9 should be near 10,000. I'm not sure about the single-core score--I've seen it range reliably between 1250 and 1350, but suspect that it might be slightly less than the i7, given the i7's higher clock speed.

 

[mlykke]

I just got clarification from Rob @ Barefeats and he confirms that the Geekbench test has been run multiple times and both Apple ram and OWC ram, but never with mixed ram. So his result of only 7321 just causes even more confusion for me, since others are saying close to 9000. Wondering what can cause such a big difference now that ram mixing isn't the cause.

 

[KrazyKanuck]

I just got clarification from Rob @ Barefeats and he confirms that the Geekbench test has been run multiple times and both Apple ram and OWC ram, but never with mixed ram. So his result of only 7321 just causes even more confusion for me, since others are saying close to 9000. Wondering what can cause such a big difference now that ram mixing isn't the cause.

Quantity of RAM also matters. My score was about that when I was just using 8GB of Apple stock RAM. It improved markedly with 32GB.

 

[mlykke]

Quantity of RAM also matters. My score was about that when I was just using 8GB of Apple stock RAM. It improved markedly with 32GB.

The Barefeats results are with 32GB.

 

[KrazyKanuck]

The Barefeats results are with 32GB.

Hmm. Not sure what to say. I just ran Geekbench again without closing any of my running applications or the podcast I am listening to and these are my results.

 

[wavesm]

In my own case I got different results with an i7; main reason is upgrading RAM with/without Apple stock ones.

Eg.

- With 8GB Apple stock I got 7492
- With only 64GB Samsung RAM I got 8781
- With Apple + Samsung (72GB) I got between ~7600 and ~8200

If you see the details in Geekbench you will find similar pattern having a look to the RAM details.

 

[KrazyKanuck]

In my own case I got different results with an i7; main reason is upgrading RAM with/without Apple stock ones.

Eg.

- With 8GB Apple stock I got 7492
- With only 64GB Samsung RAM I got 8781
- With Apple + Samsung (72GB) I got between ~7600 and ~8200

If you see the details in Geekbench you will find similar pattern having a look to the RAM details.

Indeed. This mirrors my own experience. There's something about mixing Apple RAM with 3rd party RAM in these 2020 machines. It may be mixing different module sizes, it may be something about Apple OEM (Hynix or Samsung, I think) mixing with other brands. The only reliable solution I've found so far is to remove the Apple RAM and use identical sticks of 3rd party RAM, in my case 4 sticks of 8GB Crucial Mac RAM.

 

[mlykke]

I agree that the likely issue seems to be mixing different types and that would most likely explain most of the differences on Geekbench. But when Barefeats confirm he didn't mix the ram and still got really low results, then it gets confusing.
He did say he would be testing with 64, 96 and 128 GB as well. So guess I'll await those results.

In the meantime, if anybody wants to contribute their Geekbench CPU scores along with details about which processor, ram is being used, that would be helpful. Please make sure the ram is running in dual-channel and is NOT mixed with the Apple ram.

 

[KrazyKanuck]

I agree that the likely issue seems to be mixing different types and that would most likely explain most of the differences on Geekbench. But when Barefeats confirm he didn't mix the ram and still got really low results, then it gets confusing.
He did say he would be testing with 64, 96 and 128 GB as well. So guess I'll await those results.

In the meantime, if anybody wants to contribute their Geekbench CPU scores along with details about which processor, ram is being used, that would be helpful. Please make sure the ram is running in dual-channel and is NOT mixed with the Apple ram.

Good idea. Here is another test with no other applications running and my system at idle. As mentioned before, and for reference, I'm using 4 sticks of identical Crucial 8GB Mac RAM, purchased separately in pairs.

 

[RAWvJPG]

It could be because of indexing. I guess most of these you tubers don't bother and run their benchmarks right after they set it up and they are not aware of the possibility to turn off indexing.

 

[mlykke]

Looks like an i7 with 5700 XT Pro 16GB, 10Gbit Ethernet, 1TB SSD and then upgrade with 64 or 128 GB on the side, for me. Seen a couple of video reviews that show that the i9 slows down on heavy loads compared to the i7 due to thermal restrictions.

That should tie me over until we hopefully see a re-designed 30-32" ARM iMac and the problems with the first version have been solved.

 

[wendylou]

Can you provide links to those videos? In contrast, this reviewer actually gave a positive review of the i9 overall speed:

 

[mlykke]

Can you provide links to those videos? In contrast, this reviewer actually gave a positive review of the i9 overall speed:

Sure, here's a french video that has a lot of benchmarks. In multiple tests the i7 performed better and in some the i9 was better(Usually only by a fairly small margin). Keep in mind this is if you compare the CPU. The GPU is different because the i7 in his test does not have the 5700 XT Pro.

I've added a few screenshots from the video, but I suggest to watch the entire thing to see all the different benchmarks. One of the screenshots also show the average CPU clock speed being kept higher on the i7 than the i9.

There is no question that the i9 will have a higher peak performance, but it looks like the i7 will have better single core and better sustained performance. I saw something similar in another video but I can't find it right now. I've seen way too many review videos by now 😁

 

[wendylou]

Thanks! I make multitrack music in Reason Studio so the slight single core advantage of i7 is desirable. But I’m conflicted because I would also prefer the i9 10-core multicore speed advantage for X-plane, Final Cut Pro X, and Blender 3D renderings. Any musicians out there in the same boat trying to decide which processor?
[automerge]1597675336[/automerge]